Modelling the water budget and the riverflows of the Maritsa basin in Bulgaria

HESSD4, 475–521, 2007

Modelling waterbudget of Maritsabasin in Bulgaria

E. Artinyan et al.

Title Page

Abstract Introduction

Conclusions References

Tables Figures

J I

J I

Back Close

Full Screen / Esc

Printer-friendly Version

Interactive Discussion

EGU

Hydrol. Earth Syst. Sci. Discuss., 4, 475–521, 2007www.hydrol-earth-syst-sci-discuss.net/4/475/2007/© Author(s) 2007. This work is licensedunder a Creative Commons License.

Hydrology andEarth System

SciencesDiscussions

Papers published in Hydrology and Earth System Sciences Discussions are underopen-access review for the journal Hydrology and Earth System Sciences

Modelling the water budget and theriverflows of the Maritsa basin in Bulgaria

E. Artinyan1, F. Habets2, J. Noilhan2, E. Ledoux3, D. Dimitrov1, E. Martin2, andP. Le Moigne2

1NIMH-regional centre, 139 Ruski blvd., Plovdiv, Bulgaria2Meteo-France/CNRM, 42 Coriolis ave., 31057 Toulouse, France3ENSMP/CIG, 35 St Honore st., 77305 Fontainebleau, France

Received: 8 December 2006 – Accepted: 15 February 2007 – Published: 1 March 2007

Correspondence to: E. Artinyan ([email protected])

475

http://www.hydrol-earth-syst-sci-discuss.net

http://www.hydrol-earth-syst-sci-discuss.net/4/475/2007/hessd-4-475-2007-print.pdf

http://www.hydrol-earth-syst-sci-discuss.net/4/475/2007/hessd-4-475-2007-discussion.html

http://www.copernicus.org/EGU/EGU.html

HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Abstract

A soil-vegetation-atmosphere transfer model coupled with a macroscale distributed hy-drological model was used in order to simulate the water cycle for a large region inBulgaria. To do so, an atmospheric forcing was built for two hydrological years (1 Oc-tober 1995 to 30 September 1997), at an eight km resolution. It was based on the data5

available at the National Institute of Meteorology and Hydrology (NIMH) of Bulgaria.Atmospheric parameters were carefully checked and interpolated with a high level ofdetail in space and time (3-h step). Comparing computed Penman evapotranspirationversus observed pan evaporation validated the quality of the implemented forcing. Theimpact of the human activities on the rivers (especially hydropower or irrigation) was10

taken into account. Some improvements of the hydrometeorological model were made:for better simulation of summer riverflow, two additional reservoirs were added to sim-ulate the slow component of the runoff. Those reservoirs were calibrated using the ob-served data of the 1st year, while the 2nd year was used for validation. 56 hydrologicstations and 12 dams were used for the model calibration while 41 rivergages were15

used for the validation of the model. The results compare well with the daily-observeddischarges, with good results obtained over more than 25% of the rivergages. Thesimulated snow depth was compared to daily measurements at 174 stations and theevolution of the snow water equivalent was validated at 5 sites. The process of meltingand refreezing of snow was found to be important on this region. The comparison of20

the normalized values of simulated versus measured soil moisture showed good cor-relation. The surface water budget shows large spatial variations due to the elevationinfluence on the precipitations, soil properties and vegetation variability. An inter an-nual difference was observed in the water cycle as the first year was more influencedby Mediterranean climate, while the second year was characterised by continental in-25

fluence. Energy budget shows a dominating sensible heat component in summer, dueto the fact that the water stress limits the evaporation. This study is a first step forthe implementation of an operational hydrometeorological model that could be used for

476





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

real time monitoring and forecast the water budget and the riverflow of Bulgaria.

1 Introduction

In recent years, water related problems and their management appear to be increas-ingly important in Bulgaria. This is caused partially by drought periods experiencedsince 1994, but also by the recent inundations and the economic changes. The transi-5

tion of the country towards a market economic model focuses the attention on a moreefficient water use, flood forecasting and mitigation. This increased interest requiresmore detailed and better founded information in order to provide good support for thedecision making system. The needs cover large number of fields: flood prevention, wa-ter availability for the industry, agriculture and cities, water quality management, ecol-10

ogy and climate change. Until now the water budget of the country was mostly studiedby using statistical and climatologic approaches. That made it possible to estimate thewater budget components for each climatic region of the country. The capacity of thisapproach however is too limited to offer the level of detail that is necessary for real timeevaluation of the surface and groundwater resources.15

This paper presents the first attempt to implement a soil-vegetation-atmosphere-transfer scheme (SVAT), coupled with a distributed macroscale hydrological model anddriven by observed atmospheric forcing for a large region of Bulgaria. The objective ofthis coupled model is to improve the estimation of the surface water budget (evapora-tion, soil moisture and runoff) consistently with the simulation of the riverflows. Partic-20

ularly important is to analyse the partition of precipitation into runoff and evaporationbased on a realistic description of the land surface conditions (topography, vegetation,soil).

The study is based on the application of the coupled soil-biosphere-atmosphere(ISBA) surface scheme (Noilhan and Planton, 1989) and the MODCOU macroscale25

hydrological model (Ledoux et al., 1989) which were already applied on three basinsin France: the Adour/Garonne basin (Habets et al., 1999a; Morel, 2003), the Rhone

477





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

basin (Habets et al., 1999b; Etchevers et al., 2000), and Seine basin (Rousset et al.,2004). This paper describes the first application for a region that experiences bothcontinental and Mediterranean climates, with pronounced dry period in the summer.This allows validating the functioning of the coupled model in different climatic and landcover conditions. After a description of the hydrometeorological characteristics of the5

Maritsa basin, the ISBA-MODCOU model is presented. The implementation of thehydrological model and modelling results are analysed in the two last sections.

2 Description of the Maritsa basin: hydrology and meteorological conditions

2.1 Geographic and climatic characteristics

The Maritsa river basin with its tributaries Tundzha and Arda occupies about one third10

of the surface of Bulgaria – 34 169 km2. The studied basin includes a small part onTurkish territory down to the town of Edirne, where two important subbasins (of Ardaand Tundzha rivers) reach the main river (Fig. 1) and so the total surface in ques-tion becomes 36 255 km2. Within Bulgarian borders, the river length is approximately320 km with an average slope of 7.7%. It crosses the border between Bulgaria and15

Greece and after that, until it reaches the Aegean Sea, the river serves as a naturalborderline between Greece and Turkey. Therefore Maritsa basin is an important watersource in South-Balkan peninsula, passing through three countries. The elevation ofthe Maritsa watershed goes up to 2925 m at the peak of Musala in the Rila mountain.The main geographical structures are the Thracian Valley in the centre, a part of the20

Balkan mountains (Stara Planina) at the North and the Rila and Rhodopy ranges in theSouthwest (Fig. 1). The average slope of the Bulgarian part of the basin is 12.5%.

Mediterranean influence prevails in the Southeast, where the maximum of precipita-tion comes in winter. In the central and northern part of the domain the maximum ofprecipitations occurs in May-June, due to the continental climate influence.25

Annual crops (cereals, vegetables, cotton, and tobacco) and orchards are mainly

478





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

cultivated in the valleys. The hilly areas are used as a pasture, vineyards or to cultivatepotatoes. Forests cover about 40% of the watershed surface. Oak prevails in the valleyforests while beech and pine dominate the mountain areas.

2.2 Brief description of the hydrological regime of the surface water and the aquifers

The snowfall in the mountain regions constitutes 30% to 50% of total precipitation.5

Snow cover lasts 73 to 170 days, for the Rhodopy, Stara Planina and Rila mountains(Vekilska and Kalinova, 1978). According to the climatology the mean annual waterbudget of the whole country is as follows: precipitations – 690 mm, runoff – 176 mm andevaporation 514 mm (Zyapkov, 1982). The annual averaged streamflow of the Maritsariver varies between 40 m3s−1 and 190 m3s−1 for the period from 1936 to 1975. During10

the summer, the streamflows are very low. Between July and September, the damsand the ground-water outflows mainly sustain the riverflows.

Unconfined aquifers are another specific feature of the studied area. The largeraquifer is situated in the Upper Thracian Valley. It covers an area of about 6710 km2

(Kalinova, 1982) between the three main mountain ranges. This aquifer is widely used15

for irrigation, industrial and domestic water supply. The average outflow of the mainaquifer is about 12 m3s−1 while its total storage is about 10.9×109 m3. In the valleys ofKazanlak and Sliven, other smaller basins have total reserves near about 1.14×109 m3

and 0.740×109 m3 respectively (Antonov and Danchev, 1980).Many karstic areas affect the streamflow in the Rhodopy Mountain. They have well-20

developed system of underground flow, caves and emerging springs (Fig. 2). For theperiod 1980–1996, the average discharge of the four biggest springs, Kleptuza, Beden,Devin and Tri Voditzi are respectively 0.40, 0.69, 0.53 and 1.12 m3s−1 (Machkova andDimitrov, 1990). Those perched mountainous (karstic) aquifers are partially contribut-ing to the main water table in the Upper Thracian Valley, especially from the northern25

slopes of Rhodopy Mountain. The average annual underground transfer to the plainaquifer is evaluated to 12×106 m3, while 47×106 m3 are emerging at the surface assprings (Troshanov, 1992).

479





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

2.3 Anthropogenic influence

During the years between 1950 and 1970 more than fifteen dams were built in thesouthern part of the country for better control of the riverflow. The total capacity ofthe main reservoirs is higher than 2810×106 m3 and their overall surface is larger than12 800 km2. Although few of them serve as inter-annual flow regulators, usually they5

hold the peak flow in the winter-spring seasons and release water to produce energyand for irrigation in the summer. Most often the dams are built on the riverbed but infew cases, they are in derivations. Figure 2 shows the location of the main dams in thebasin. Other main anthropogenic influence is the direct use of water from the river forirrigation. There are also cases of transferring water from one river basin to another.10

For example, after the Koprinka dam on the Tundzha River, a catchment takes water forirrigation and hydropower producing purposes and transfers it into the basin of Sazliikariver that is not tributary of Tundzha River. Figure 3a presents the effect of differentcases of anthropogenic influence on the riverflow for the period 11/1995 to 10/1996.The most important impact is a result of the dams’ water storage and release. From15

June to November the dams contributions represents from 1% to 33% of the riverflowwhile in January, February and April they store 25 to 27% of the streamflow. Althoughtheir annual balance is compensated, they play a considerable role in the monthly par-tition of the riverflow. For the same period, the water used for irrigation is about 12 mmthat is almost 7% of the riverflow. The amount of water transferred from other basins20

is about 4% of the streamflow. The overall monthly impact of human influence on thenatural streamflow for the period 11/1995 to 10/1996 is shown in Fig. 3b. Section 4.2provides details on the implementation of the anthropogenic influence in the simulation.

480





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

3 The ISBA-MODCOU model

3.1 Description of ISBA land surface scheme

The ISBA surface scheme was developed for the climate, mesoscale and predictionatmospheric models used at Meteo-France. It represents the main surface processesin a relatively simple way: it solves one energy budget for the soil and vegetation5

continuum, and uses the force-restore method to compute energy and water transfersin the soil. The two soil layers representation is used – a shallow surface and a rootzone (Fig. 4). Four components are used to compute the evaporation: interception bythe foliage, bare soil evaporation, transpiration of the vegetation, and sublimation of thesnowpack.10

Two fluxes of water in the soil are computed: a surface runoff (Qr ) and drainage(D)(Fig. 4). Subgrid heterogeneities of the soil moisture are involved only in the surfacerunoff. To evaluate the subgrid runoff, the concept of the Variable Infiltration Capacity(VIC) (Dumenil and Todini, 1992) is used. It considers that a fraction of the cell is sat-urated, and thus, can produce surface runoff, even if the whole mesh is not saturated.15

Such fraction is almost zero when the soil is dry (around the wilting point), and is goingup to 100% when the whole cell is saturated. It is varying according to an exponentialfunction, which is based on a shape parameter (b).

In this application, the snow pack is represented by one layer with uniform tempera-ture, density and water content (Douville et al., 1995).20

3.2 Description of the conception of additional reservoirs for the drainage flow

The previous applications of the coupled ISBA-MODCOU model were done for rel-atively wet regions, without pronounced dry periods. In the case of the Maritsa riverduring the dry period of the year the runoff is mostly sustained by the deep soil drainageand the water table, where it exists. The process occurring in the unsaturated zone,25

between the soil root-zone and the water table, has a high contribution to the total

481





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

runoff especially in the summer. In the two-layer soil scheme, used in this application,water transfer in the unsaturated zone was not implemented. For better simulation ofthe observed time delay in the streamflow, two additional reservoirs were introduced,between the surface scheme ISBA and the MODCOU hydrological model (Fig. 4). Inthe mountain area, where aquifer layer does not exist, only these additional reservoirs5

simulate the time lag of the drainage water during the transfer in the unsaturated zone.In this new module, a fraction α of the gravitational drainage simulated by ISBA (D)is transferred to the first reservoir. This reservoir has a water content h1 and a deple-tion coefficient C1 that is relatively low in order to induce large time delay. When itsmaximum level h1max is reached, water surplus is transferred to the second reservoir.10

This second reservoir represents the less compact and more fissured upper area ofthe geological profile. Therefore it has a higher depletion coefficient C2, leading to ashorter time delay. When the second reservoir level h2 reaches its maximum – h2max,the extra water leaves the reservoirs. Thus, the drainage part of the runoff is formedby Eq. (1) and (2):15

Qd = D × (1 − α) +Qov + h1 × C1 + h2 × C2 (1)

With Qov = D × α − (h2 max−h2) (2)

Therefore the five parameters of the additional reservoirs are:α – coefficient controlling the drainage water input into the reservoirsh1max, h2max – maximum levels of the reservoirs20

C1, C2 – depletion coefficients of the reservoirs, C1≤C2Qd is transferred either to the riverflow where there is no aquifer or to the water

table simulated by MODCOU. The part of it that passes through the first reservoir(h1×C1)could be considered as the slower part of drainage. The additional reservoirsare processed at the time step and using the grid of the hydrological model. The higher25

resolution of the hydrological grid gives the possibility to calibrate the parameters bynested subbasins, which is described in Sect. 5.2. Parameters variability could be

482





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

related to the geomorphic characteristics (elevation, slope) or to the geologic profile, ifsuch information is available.

3.3 Description of MODCOU

The macro-scale hydrological model MODCOU was used in various applications(Ledoux et al., 1989). MODCOU takes into account the surface and underground5

layers. The surface routing network is computed starting with the topography, by usinga geographical information system. The surface and underground domains are dividedinto grid cells of embedded size (from 1 to 4 km), the higher resolution being associ-ated to the river grid cells. The transfer time between two grid cells is based on thetopography, the distance between the cells and the surface of the basin. The surface10

runoff computed by ISBA is routed to the river network and then to the river gages us-ing isochronous zones with a daily time step. The drainage computed from ISBA andfrom the new drainage module contributes to the evolution of the groundwater table,which evolves according to the diffusivity equation. Exchange of water between thegroundwater table and the river are computed according to simple relations (Ledoux et15

al., 1989). At the end, the flows from the surface layer and from the groundwater tableform the riverflow at the gauging stations.

4 Implementation of ISBA-MODCOU in the Maritsa basin

4.1 Hydrological parameters

The hydrographical surface network as well as the underground layer grid was estab-20

lished by using a GIS based on the topography (Golaz et al., 2001). For that purpose,the GTOPO30 database (provided from USGS EROS Data Centre) was used. Thegrid consists of 11 661 meshes in the surface layer, including 2387 river cells; and4390 cells for the underground layer (Fig. 2).

483





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

A maximum transfer time Tc for the water to reach the outlet was established forthe Maritsa basin, according to the observed streamflow: Tc=6 days. The evolution ofgroundwater table is controlled by transmissivity and storage coefficients. They werecalibrated for eight subregions of the unconfined underground layer. The existing pub-lications were used to estimate a first guess of these coefficients. Transmissivity varies5

from 1.0×10−3 to 34×10−3 m2s−1 while the values of the storage coefficient are be-tween 0.20 and 0.23 (Antonov, 1980).

The five additional parameters of drainage reservoirs had to be calibrated for eachof the 68 subwatersheds

4.2 Implementation of dam reservoirs in the simulation10

Data about the water budget of twelve reservoirs and about water redirecting and chan-nelling were collected for the first year of simulation. They were used for the calibrationperiod, first, to validate the simulated streamflow at the dam entrance and, second, toimpose the dam release to the simulated streamflow after the dam.

In the simulation, all the streamflows that are downstream the dams or the redirecting15

points were corrected in order to take into account the impact of the constructionsalong the riverflow. This correction was achieved by considering the time lag impliedby the storage and the transfer in the channel. Natural riverflows of Maritsa river (8gages), Tundzha river (2 gages), Chepinska river (2 gages), Vacha river (4 gages), aswell as the Arda river outlet were corrected, by subtracting the simulated natural flow20

coming from the rivercells just upstream the dams, and by adding the observed dams’water release while respecting the time lag between the two cells. In the northernpart of the basin, the observations showed that the outflows from irrigation dams didnot sustain the riverflow in summer. Instead of that, those outflows were redirectedthrough irrigation channels (Fig. 3a). Therefore, this part of the streamflow (11.7 mm25

or 7%) was subtracted from the simulated streamflow. At the basin level, the overalleffect of anthropogenic influence for the period 1 November 1995 to 30 October 1996was about −4 mm, and near 7 mm was transferred from other basins (Fig. 3a).

484





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

4.3 Surface parameters

The ISBA parameters can be determined by the soil texture and the vegetation mapsusing tables of correspondence, as detailed in Noilhan and Lacarrere (1995).

The vegetation map compiled by Champeaux and Legleau (1995) from the NDVIarchive distinguishes 12 vegetation types. The resulting vegetation map (Fig. 5) shows5

that forests are the dominant vegetation type in the mountains. In the valley, Crop(3) – that is interpreted as Mediterranean region cereal, associated with dry summerconditions, is dominant. The Rock type influences few grids in the high mountain areaof Rila and Stara Planina. A single vegetation class stands for any forest type. Themonthly evolution of leaf area index (LAI), vegetation cover (VEG), and roughness10

length (z0v ), were related to the 2 years’ satellite archive of the advanced very highresolution radiometer/normalized difference vegetation index (AVHRR/NDVI), followingthe method presented in Habets et al. (1999a). The minimum surface resistance (Rsm)and albedo (αv ) are constant in time and linked to the vegetation type.

In the study detailed maps of the soil properties - the percentage of sand and clay15

as well as the soil depth linked to the maximum depth of the root system of cultivatedcrops by agricultural region were established (Trendafilov, 19961). These maps wereused to obtain data at 1 km resolution. The soil depth map was compared to the soildepths derived from the vegetation type. The last one gives more than 150 cm depthfor the forested area, which is not realistic for the mountain forests in Bulgaria (Ninov,20

1982). The soil depth used in this study varies between 40 cm and 150 cm. Only theforested regions in the valley have deeper soil – 180 cm.

The calibration of the b parameter used in the subgrid runoff scheme was performedby using the same ideas as in Dumenil and Todini (1992). It depends on the altitude,

1Trendafilov, Ch.: Maps of soil mechanical properties in the region of Maritsa river basin,personal archive of the author, 1996.

485





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

as presented in Eq. (3),

b =(

0.2 +Alti3000

)× 1.4, (3)

with Alti is the cell elevation [m] and 0.28≤b≤1.68.Equation (3) gives the best modelling results for high flow conditions. The values

of b are significantly higher than the values calibrated in preceding application (Ar-5

tinian, 1996). This result is explained with the introduction of additional reservoirs forthe drainage flow, as now the efficient simulation of peak flows needs lowering of theprecipitation’s fraction transferred to drainage reservoirs.

4.4 Atmospheric forcing

4.4.1 Meteorological database10

To compute the water and energy cycle the ISBA surface scheme needs 8 atmosphericparameters: rainfall and snowfall, air temperature and humidity at 2 m, wind velocity,atmospheric pressure, global and atmospheric radiations. For the application overFrance, the SAFRAN analysis system is used (Quintana Seguı et al., 20072). Suchanalysis system was not yet implemented in Bulgaria, and thus, an important work was15

done in order to generate the atmospheric database.Such database was assembled for 26 months (from 1 August 1995 to 30 September

1997). The following data sources were available at the National Institute for Meteo-rology and Hydrology of Bulgaria (NIMH): 12 synoptic stations, recording atmosphericparameters each 3 h; 55 climatologic stations with 3 values a day – at 07:00, 14:00 and20

21:00 h LT and 175 precipitation stations – measuring the daily precipitation and snowdepth (Tables 1 and 2). For the first year snow density data of five additional stations

2Quintana Seguı, P., Le Moigne, P., Durand, Y., Martin, E., Habets, F., Baillon, M., Fran-chisteguy, L., Morel, S., and Noilhan, J.: The SAFRAN atmospheric analysis, Description andvalidation, Journal of Applied Meteorology and Climatology, under review, 2007.

486





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

were available. The global solar radiation data were obtained from two stations, one inthe valley close to the town of Chirpan and the second one in a mountain location at1800 m a.s.l. Data with 3-h step was taken from monthly paper reports. The continu-ous records of global radiation of Chirpan station were obtained on graph paper strips– one per day. These analogous records were scanned and hourly sums were com-5

puted by integration. The atmospheric forcing was prepared at a 3-h timestep; whilethe precipitations were collected on a daily basis. To be consistent with the densityof the observation network and the hydrological grid, an 8×8 km grid cell was used tointerpolate in space the atmospheric forcing. This meteorological grid consists of 638cells (Fig. 2). Next two sections present the preparation of the atmospheric parameters10

needed for the modelling.

4.4.2 Snow and rain precipitation, air temperature, wind velocity and specific air mois-ture fields

In order to select only realistic data, a criterion based on the standard deviation (σ) wasused to isolate erroneous data. It reflects the variability of a parameter around its aver-15

age value for certain periods so that errors in data series of relatively high homogeneity,as air temperature, relative humidity and wind velocity should be easily detected. How-ever, it can’t be used to estimate the errors in precipitation data series. Air temperature,wind velocity and relative moisture records were carefully checked and corrected usingthe ±3×σ rejection criterion. The precipitation data were checked by comparison with20

the climatological maps of precipitation for a given season (Hershkovich et al., 1982).The point scale observations were interpolated in space using two software pack-

ages. The spatial interpolation of the temperature was made with software dedicatedto scattered data, statistically linked to the topography, the Aurelhy method (Benichouand Le Breton, 1987). However, when the temperature field is not enough correlated25

to the elevation, due to atmospheric temperature inversion that appears most oftenin winter time, the krigging software Bluepack was preferred. This method leads toacceptable results, with higher quality than Aurelhy.

487





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

The temperature observations come from two meteorological networks with differenttime steps of observations (climatological and synoptic). The first one has higher detailin space, while the second is more detailed in time. To produce temperature forcingwith good resolution both in space and time, we combined the two fields in one newfield using a spline function that approximates the daily temperature variability. The5

same method was used to work out the field of the relative air moisture, needed tocompute the specific air humidity. The other atmospheric fields (wind velocity, rain andsnow precipitation) were interpolated with the Bluepack krigging software.

The interpolation of precipitation is difficult because of its high spatial variability.Where the rain gauges are too close, the Bluepack krigging method gives noisy results.10

Noise analyse showed that it depends on the average distance between two stationsfor the whole field. That brought the idea of “averaging neighbours” method: where thedistance between two precipitation stations is less than the required minimum, the av-erage of the observed values of the two gauges is attributed to both stations before theinterpolation. For the Rhone basin precipitation field, the minimum tolerable distance15

was 2 km (Artinian, 1996). In the case of the Maritsa river basin, where the rain gaugesare scarce, this distance was determined to be 6 km.

Surface atmospheric pressure was estimated directly from the elevation because thevariability due to the topography is several times higher compared to the seasonal one.

Specific humidity of the air at 2 m was calculated using values of the atmospheric20

pressure, temperature and relative humidity of the air.

4.4.3 Atmospheric radiation and global radiation fields

To compute the atmospheric radiation the formula of Staley and Jurica (1972) wasused. It takes into account the air temperature, air specific humidity and cloudiness.

For the global radiation, few observations were available: only two stations had mea-25

sured directly the global radiation at a hourly time step: Chirpan in the plain (173 m)and Rozhen in the mountain (1750 m). Measurements of the bright sunshine hourswere made at 15 sites, using a sunshine recorder, which allows the estimation of the

488





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

“bright sunshine ratio” i.e., the ratio of the actual bright sunshine to potential brightsunshine. Also, cloudiness was observed at 55 sites.

Global radiation depends on the elevation, because of the impact of the aerosolconcentration on the atmospheric transmittance (Hottel, 1976) and air turbidity. To takeinto account such impact, as well as all the available data, we used a modified version5

of the parameterisation suggested by Kasten and Czeplak (1979). It is based on astatistical relation between the hourly global radiation, the hourly bright sunshine ratio,and the 3-hourly cloudiness, observed at the different sites. It is expressed as follows(Eq. 4)

Rg = Rg0 ×{A1 ×

(1 − 0.88

(Nb10

)3.5)

+ B1 ×Alti1000

+ C1 × Sun

}(4)10

where Rg stands for the hourly global solar radiation (Wm−2), Rg0 stands for theoreticalclear-sky global radiation computed according to the solar elevation angle at sea level(Kasten and Czeplak, 1979), Nb stands for the average hourly cloudiness (varyingfrom 0 to 10), Alti stands for the altitude of the grid point (m), Sun stands for the hourlybright sunshine ratio, ranging from 0 to 1 and A1, B1, C1 – are empirical coefficients15

The empirical coefficients A1, B1 and C1 were found to depend on the value of Rg0,so they were established for 10 intervals of Rg0. These coefficients, when computedfor the entire range of Rg0 between 0 and 900 Wm−2 were: A1=0.288, B1=0.196 andC1=0.691.

The atmospheric parameters were computed for two hydrologic years – 1995/199620

and 1996/1997. The maps of the atmospheric forcing for the two years are presentedin Fig. 6. It shows the annual accumulated total precipitation, snowfall, mean annual airtemperature and global solar radiation for the two years of simulation. There is a largespatial variability. The total precipitation is marked by strong values in the mountainareas, but the higher values for the second year are situated in the Southeast, where25

the more pronounced Mediterranean climate can produce intense rainfall. In the valleythe annual value of total precipitation varies from 400 to 600 mm, while in the mountains

489





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

it varies from 700 to 1300 mm. Snowfall is more important in the mountain. It variesfrom 10 mm in the valley to 800 mm in the Rila Mountain.

4.4.4 Validation of the atmospheric forcing

As all the available data were used to establish the atmospheric forcing, it is not pos-sible to validate it directly. However, indirect validation can be made, by using pan5

evaporation observations. Pan evaporation depends on the atmospheric conditions.It can then be compared to pan evaporation computed from the new data set. Suchcomparison relies on the hypothesis that the pan evaporation is comparable to thePenman potential evapotranspiration. The accumulated Penman evaporative demandof the atmosphere (Choisnel, 1988) was computed, at a 10-day timestep, by using the10

interpolated values for temperature, radiation and wind speed, and compared (Fig. 7)to the pan evaporation observed at 5 sites. The coefficient of determination of thecomparison is R2=0.87 and the root mean squared error RMSE=5.3 mm.

5 Modelling results

5.1 Hydrological database and methodology15

The hydrological database consists of daily streamflow discharge of 56 river gages.The inflow and release flows data of 12 dams, as well as snow density measurementsonly for the first year from the National Electricity Company (NEK ); and from the WaterManagement Company (Vodno Stopanstvo) for the northern part of the basin were ob-tained (Table 2). From the total number of 68 stations and dams, only 41 could be used20

for statistical comparisons, because for 10 stations and 12 dams no data was availablefor the validation year. Additionally, five small catchments (smaller than 50 km2) werediscarded from the comparison because there was a 10% error between the reportedsurface of the subbasin and the modelled one. In order to check the quality of the

490





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

simulation, three statistical criteria were computed for each gauging station: the ra-tio between simulated and observed annual discharge Qsim/Qobs, the daily efficiencyE (Nash and Sutcliffe, 1970), and the coefficient of determination – R2. To achieveperfect simulation these three statistical numbers should be equal to 1.0. The ratioQsim/Qobs gives an estimation of the annual partitioning of the precipitation into runoff5

and evaporation, whereas R2 indicates if the simulated and the observed streamfloware significantly correlated. The efficiency E is an intermediate criterion, very sensitiveto the flood overestimation.

In the following sections, the method of calibration of the unsaturated reservoirs ispresented, and the results in terms of streamflow, snow height, soil moisture, and water10

and energy budgets are discussed.

5.2 Calibration of the unsaturated zone reservoirs

The calibration method is a multiple step optimisation procedure based on the sta-tistical results of the comparison between simulated and observed streamflow. First,the extreme limits of the parameters were set. For each subbasin, the total volume15

of the runoff for the dry period of the year – Qdry [mm] was computed. This firstguess value is assumed to be close to the average level h1 of the first reservoir. Theinitial estimation of the extreme values for the parameter h1max where chosen to be10×Qdry≥h1max≤ 1

4×Qdry. The limits for the depletion coefficient C1 were deductedfrom extreme values of h1max in order to simulate the average daily streamflow dur-20

ing dry periods. The parameter h2max was initialised with the same extreme levels.To initialise the parameter C2, we used the relations C1<C2 and C2<0.20, as with thevalue of 0.2 a reservoir of h2max=300 mm (maximum value found in the previous step)is depleted in about 5 days. The parameter α varies between 0.0 and 1.0. An iterativeprocedure was undertaken, with cycles between the extremes for each parameter, us-25

ing a large step. Each time when the statistics were higher than in the previous step,the resulting combination of parameters was stored. This procedure was repeatedwith a smaller step by using the parameters already defined in the previous step for

491





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

the initialisation. The optimisation process was repeated until the statistics achievedconvergence. Table 3 shows the computed average and extreme values of the fiveparameters.

At the end of the calibration phase, the following results were observed:

– The valley subcatchments present high values of the coefficient: α=0.80 to 1.0,5

which means that only small part of the streamflow is a rapid flow.

– The Southeast part of the basin (Arda and tributaries) shows low rate of drainagewater storage: α=0.05 to 0.35.

– The subcatchments from regions with pronounced karst development in theRhodopy Mountain show higher α coefficient: α=0.75 to 1.0 than the other catch-10

ments with the same average elevation (α=0.1 to 0.65).

These observations correspond to some published results about the partition of therunoff according to its origin – surface, drainage and deep drainage (Yordanova, 1978).However, in many cases the lack of knowledge about the anthropogenic activity leads toerrors during calibration. In order to estimate the parameters with high quality, detailed15

information about human activity in the studied area is necessary.

5.3 Results in term of streamflow simulation

For the entire studied area, the error on the mean annual discharge for the first (calibra-tion) year is lower than 20% for half of the stations. On average, the annual simulationfor the calibration year is close to the observations for the main stations (overestimation20

of 13%). The observed and simulated daily streamflow discharges for 8 river gages aregiven (Fig. 8). Their positions are shown in Fig. 2.

The value of E is greater than 0.7 for 27% of the stations and greater than 0.6 for 36%of them. The best values are obtained for the main rivers (Fig. 8a to d and Table 4).

For the calibration year, as it could be expected, better results are obtained when25

the dam inflow/outflow are taken into account (Table 4), except for the annual ratio492





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Qsim/Qobs that is overestimated. This overestimation is due to the underestimation ofthe fraction of the dam water release that is used for irrigation purposes. This amountshould be removed from the simulated riverflow. The influence of imposed streamflowis stronger near to the dams (for instance at Plovdiv) and diminishes downstream (forinstance at Svilengrad). It rises again at the outlet (Edirne) because of the proximity to5

the Arda river reservoir cascade.The lowest efficiencies are computed for the Northwest part of the Tololnitza and

Striama watersheds, where the rain gauges are too few.For the second validation year (1996/1997) the efficiency is lower. The value of E is

higher than 0.6 for 32% of the stations, but the error on mean annual discharge remains10

at the same level – lower than 20% for half of the stations. Twelve gauging stations,not perturbed by dams, have higher statistic results for the validation year than for theyear of calibration.

5.4 Snow simulation

To validate the snow cover evolution simulated by the ISBA surface scheme, 20 clima-15

tological stations were selected according to the following criteria: minimum elevation450 m, more than 100 days of observed snow cover for the two years, observationsavailable for the entire simulation period and grid cell altitude close to the station’s ele-vation (difference lower than 200 m). At these sites, the interpolated air temperature iscloser to the observed one comparing to sites with only rain/snow measurement sta-20

tions. The mean evolution of snow depth for these stations is depicted in Fig. 9a and b.The second year, the results are poorer and one of the reasons for this is the generallyhigher temperature in the winter of 1997 (on average for January 1996 it is −1.64◦C;while for January 1997 it is +2.01◦C). Therefore the melting of snow pack during theday and the refreezing of the liquid water stored in the snow pack at night happens25

more often during the second year than the first one. Such a process is not taken intoaccount in the one-layer snow scheme used in this study. It was however simulated bythe 3 layers snow schemes recently developed for ISBA (Boone and Etchevers, 2001),

493





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

and the application of that scheme showed good results for both years. The RMSE forthe daily snow depth for the first year of the simulation for these sites is 5 cm but for thesecond simulated year it is 13 cm. Observations of the snow pack from 174 stations ata daily step were considered in order to evaluate the quality of snow simulation at thebasin-range. A comparison between the averaged observed and simulated snow depth5

for 174 stations is presented in Fig. 9d. This scatter plot shows the snow scheme effi-ciency but the result is highly influenced by the quality of air temperature interpolation,which is steady in the neighbourhood of the 55 climatological stations.

Snow density together with snow depth data is available for five sites and only for thefirst modelled winter – 1995/96. The snow density and snow water equivalent (SWE)10

were compared after averaging the results of corresponding model grid cells and thedaily data of five measuring sites (Fig. 9c).

The model simulates well the snow height and the corresponding water content incold conditions (T ◦C<0.0). In case of rainfall over the snow pack (T ◦C>0.0), and peri-ods of melting-freezing, the SWE is strongly underestimated. This is leading to a lower15

SWE, than the observed one, during the less cold period. As mentionned the snowscheme efficiency was improved with the 3-layer snow scheme of ISBA developmentsnot used in that study.

5.5 Soil moisture simulation

In order to validate the model soil moisture simulation agro-meteorological data were20

collected from 10 stations measuring the soil water content (Table 5).The measurements of volumetric soil moisture were available from ten agro-climatic

stations, each one with three profiles. Agrometeorologists systematically selected thethree profiles of each site with different vegetation cover – one with wheat or barleyvegetation (winter crops), one with perennial vegetation (ex. vineyard, rose, etc.) and25

one with annual vegetation (cotton, lucerne or corn). The measurements were madeby weighting the soil sample, extracted three times a month, before and after drying.No measurements were made during the winter season. The soil and vegetation char-

494





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

acteristics observed in situ and those used in the simulation for the ten sites are givenin Table 5.

The soils in Bulgaria’s valley have often high available water capacity (AWC)(Richards and Wadleigh, 1952). Alluvial meadow soils, for instance, contain highamount of clay – between 40 and 60% and a corresponding high AWC e.g. 110 to5

180 mm for the top 1 m soil depth (Dimitrova, 1991), especially where the humus con-tent increases. AWC computed by ISBA (75 to 90 mm) are lower than the measure-ments. To compare the evolution of the observed against the simulated soil watercontent a normalization of both values was made. Thus, the moisture computed by thesurface scheme is normalized by using Eq. (5) and then compared to the normalized10

measured top 1-m soil moisture. In the last two columns of Table 5 the statistics of thecomparison are given.

Wn =w2 − w2 min

w2 max−w2 min(5)

where w2min and w2max are the minimum and maximum values of the compared gridcell or respectively the observed soil profile and w2 is the actual soil volumetric water15

content (m3/m3). The analysis of the measured values showed that all profiles of somesites were highly influenced by irrigation, so these stations were discarded from thecomparison. As the winter cereals are less dependent on water supply, they are usuallynot irrigated. Thus, two kinds of validation were made: the first one used the averagedvalues of all the soil profiles of the seven not-irrigated sites (21 profiles) and the second20

one – only the cereal profiles of these sites, i.e. wheat and barley. The comparisonshowed that for most profiles, the higher and lower values of observed and simulatedsoil moisture had a good correlation in time. The model simulates fast lowering of thesoil moisture in April and May, which corresponds to the seasonal behaviour (in termsof soil moisture usage) of the observed winter crops. The results proved that winter25

cereals observed in situ were properly defined (in terms of soil moisture usage) by theprescribed vegetation types – Crop (3) and Crop (4), except for the period from Juneto August when soil moisture was depleted faster in the simulation. Too low prescribed

495





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

LAI and VEG for those months could be the reason for this overestimation of the baresoil evaporation (LEG) (Fig. 11b).

5.6 Water and energy budgets at the basin scale

5.6.1 Surface water budget

Figure 10 shows the annual maps of accumulated evaporation and runoff for the two5

years of simulation. The fields have large spatial variability.The total evaporation is linked to the topography. Accumulated annual evaporation

varies from 300 to 780 mm. The highest value (780 mm) is simulated the second yearin the Rhodopy Mountain. The generally higher values in the mountains are related tothe dense forest vegetation, and the more important rainfall at this altitude.10

The annual accumulated runoff varies spatially from 15 mm to 580 mm for the firstyear. During the second year it ranges from 12 to 680 mm. The valleys show thelowest values, while the region of East Rhodopy Mountain is producing systematicallythe higher runoff. This phenomenon is linked to the combination of several factors:almost no forest, shallow soil and the intense rainfall. While in the forested areas15

rainfall is subject to retention by the forest litter and the evaporation rate can be high,for the south-eastern part of the Rhodopy mountain flash floods are occurring almostevery year. The drainage fraction represents 77% of the runoff for the first year and74% for the second one – respectively 125 and 126 mm (Table 6). However, almostall of it comes from the mountain area. The drainage in the valley remains very low20

because of the high evapotranspiration and its contribution to the aquifer is weak. Anexception of that could be the deep infiltration fraction of water used for irrigation.

Monthly values of the water budget (Fig. 11a) show that there are three precipitationmaximums during the first year: in November-December, in February and in Septem-ber. For the second year the maximums are in November, March–April and August.25

The first year is dominated by the Mediterranean climate (with winter precipitations)while the second year is typical for the continental water cycle. The evaporation follows

496





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

the temperature variability and the water availability. Thus, it has higher values in Apriland May (87, 84 mm), for the first year, when the two conditions intervene. It is alsolinked to the development of vegetation. It rises also in September (61 mm) with theincrease of precipitations. The bare ground evaporation (LEG) causes the Septem-ber rise while in the other cases (Fig. 11b) the plant transpiration (LETR) represents5

the larger fraction of the total evaporation. In summer, transpiration is lower (30 and20 mm) because of the water stress. For the second year the total evaporation (LE)is higher by about 100 mm (Table 6), caused by the spring and summer precipitationstogether with the higher temperatures in that season. The runoff variability is linkedto the same processes. When the precipitations occur in winter, they contribute to the10

runoff because of the low evaporation. In the opposite, huge part of the spring and allthe summer precipitations evaporate and do not contribute to the runoff.

The soil water content rises for the first year between October and December, thenin February and decreases very fast from March to May (Fig. 11c). For the period fromJune to September the soil reservoir water content rises with about 100 mm. For the15

second year the process of replenishment is shorter but more intensive in winter. Thedepleting occurs one month later due to the spring precipitations. Figure 11c showsthat the unsaturated zone reservoir plays an important role on the water budget as well.

The water budget is highly influenced by the contrast between the valley and themountain areas of the region. However, part of this contrast is hidden by the dams’ im-20

pact on the runoff. Mountainous catchments are highly influenced by the snowmelt. Onthe opposite, Arda river and its tributaries, which are under Mediterranean influence,are not affected by snow. Table 6 shows the components of the water budget of fourmain subbasins and also of four smaller watersheds not disturbed by anthropogenicactivity. For the entire basin the relation between evaporation and precipitation (E/P)25

is about 0.7 for the first year and 0.8 for the second. For comparison the mountaincatchments (Fig. 8g) have values between 0.6 and 0.65, which is due to the snowmeltfeeding and low temperatures. Southeast Rhodopy Mountain tributaries – Vurbitza andKrumovitza rivers (Fig. 8e and f), show lower values of that parameter: 0.43 and 0.51.

497





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

The shallow soils, intensive precipitations and the lack of forests in the region explainthis phenomenon.

5.6.2 Aquifer water budget

At the basin scale, the water table (plain aquifer) has a relatively small contributionto the runoff. That is partially due to the small amount of infiltration in the area with5

aquifer layer. As that layer covers the valley where the evapotranspiration takes largepart of the precipitations, only small part of the infiltration water reaches the watertable. The aquifer maintains the riverflow with 19 m3s−1, or 6% of the total streamflow,which corresponds to the reported data (Antonov and Danchev, 1980). The rechargeoccurs mainly in winter and spring months at a rate of 6–7 mm per year. The monthly10

aquifer budget is positive only for March 1996. The two-year’s budget is negative, about−10 mm. This corresponds to a decrease of aquifer level. Recharges by infiltration ofirrigation water and lateral underground recharge are not taken into account.

5.6.3 Energy budget

The energy budget is linked to the water budget by the evapotranspiration term and is15

expressed by the Eq. (6):

Rn = H + LE + G (6)

where Rn stands for the net radiation flux, H and LE stand for the sensible and latentheat fluxes and G stands for the ground heat flux. The annual variations of these fluxesare driven by the net radiation flux. The monthly budget of the studied area, for the two20

years of simulation, is presented in Fig. 12a. Rn varies between 15 and 130 Wm−2.The higher values are in June–August – over 115 Wm−2 for the two-year simulation,while the lower values are in the period November–February – below 25 Wm−2. Theevaporation fluxes varies from 8 to 70 Wm−2 between the winter and spring months.The higher values are in spring because the evaporative demand of the atmosphere25

498





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

coincides with the water availability in the soil. In summer, the latent heat is lowerthan the sensible heat because of the lack of water for evaporation. During the secondyear, more water is available in summer and the two main components of the energybudget are closer. The ground heat flux has its maximum values in March–May –4.3–4.5 Wm−2. The dominating sensible heat flux in summer months is due to two5

main reasons: the lack of precipitation and the prescribed vegetation type with lowvalues of the vegetation fraction in summer. This leads to heating of the bare groundand consequently with a water stress to reduce evaporation, and increase the sensibleheat flux. The Bowen ratio (H/LE) for the first year is equal to 1.36 and to 0.92 forthe second one. The simulated evolution of the energy budget components is close10

to the published values for the whole country territory (Vekilska, 1982). The publishedclimatological values of ratio LE/Rn vary between 0.42 and 0.70 (in average 0.46 inthe simulation) and the relation H/Rn is between 0.30 and 0.45 (in average 0.52 in thesimulation). Finally, the ground heat flux is positive from March to August. Figure 12bshows the above mentioned published values, converted into Wm−2, compared to the15

modelling results.

6 Conclusion

The purpose of this project was to implement a coupled hydrometeorological model(ISBA-MODCOU) in Bulgaria, in order to study the variability of water and energy bud-gets.20

The hydrometeorological model was already used in France, in association with theSAFRAN atmospherical analysis system. As such system was not available in Bulgariacomprehensive work was done to generate a complete atmospheric database. It hasbeen demonstrated that even with the relatively scarce meteorological network, theavailable data is qualitatively and quantitatively sufficient for the modelling. However, a25

huge preparatory work was needed, in order to extract data from various formats, oftenon paper support, to correct and interpolate the point scale observations.

499





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

In order to improve the simulation of the riverflows, a simplified scheme describingthe impact of the unsaturated zone was added. It consists of two reservoirs fed by thedrainage simulated by ISBA and allows simulation of the time delay for the transfer ofwater from the soil column to the aquifer or the river. As those reservoirs use the samedaily time step than the hydrological model, their five parameters could be calibrated5

versus the observations, with high spatial resolution and short computing time.The impact of the numerous dams and pumping the river was quantified thanks to

the numerous data collected. Such impact has a clear annual cycle, and, for a givenmonth, it can represent up to 4% of the annual discharge. Due to the precision of thedata collected, the effects of the dams on the riverflow could be taken into account in10

the simulation.The simulation was made over two annual cycles, and the validation was performed

by using observed snow depth, snow water equivalent, daily riverflow, and soil mois-ture. The simulation was in good agreement with the observations. For instance, morethan 25% of the rivergages were simulated with efficiency above 0.7.15

The study shows that the country experiences water stress in summer, which limitsthe evapotranspiration. Indeed, the annual Bowen ratio is rather high −1.36 in 1995–1996 and 0.92 in 1996–1997.

The results of this first application for the Maritsa, Tundzha and Arda basins in Bul-garia will be used in many directions:20

– It’s a first step for the implementation of an operational hydrological model thatcould be used for both monitoring and forecast of water budget and riverflow.This is a priority after the inundations in August 2005 and March 2006. Theseevents lead to economical losses of more than 850×106€ only in Bulgaria. Intime of floods, after crossing the Bulgarian territory, Maritsa and its tributaries25

Arda and Tundzha rivers cause inundations in Turkey and Greece. Therefore theimplementation of an efficient operational hydrological forecasting system in theregion will have highly positive cross border impact.

500





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

– It allows to define the methodology and to estimate the amount of data neededfor a long-term retrospective study. Such study is necessary to understand thecharacteristics of the water system in Bulgaria, and to be able to anticipate theimpact of climatic change.

– To optimise the meteorological and hydrological network in order to reduce their5

maintenance cost since financial resources for public domain are a major issue.

Acknowledgements. The authors wish to acknowledge the financial support of Meteo-Franceand Ecole Nationale Superieure des Mines de Paris (ENSMP). The authors would like to ex-press their deep gratitude to all colleagues from GMME/MC2 – division of the National Centrefor Meteorological Research of Meteo-France (CNRM), to T. Bojkova and the colleagues from10

NIMH for their suggestions and their help to obtain and prepare the necessary data sets for theperformance of this study.

References

Antonov, Ch. and Danchev, D.: Underground waters in Popular Republic of Bulgaria. StatePublishing House: Technika, Sofia 1980, 1980.15

Artinian, E.: Modelisation hydrologique du bassin du Rhone. ENM, Meteo-France, Rapport demastere, decembre 1996, Toulouse, 1996.

Benichou, P. and Le Breton, Od.: Prise en compte de la topographie pour la cartographiedes champs pluviometriques statistiques, in: La Meteorologie, 7e serie, n 19, octobre 1987,1987.20

Boone, A. and Etchevers, P.: An intercomparison of three snow schemes of varying complexitycoupled to the same land surface model: local-scale evaluation at an alpine site, Journal ofHydrometeorology, 4, 374–394, 2001.

Champeaux, J.-L. and Legleau, H.: Vegetation mapping over Europe using NOAA/AVHRR, in:The 1995 meteorological satellite data user’s conference, pages 139–143, Winchester, UK,25

EUMETSAT, 1995.Choisnel, E.: Estimation de l’evapotranspiration potentielle a partir des donees

meteorologiques, La Meteorologie, 7, 19–27, 1988.Dimitrova, Yu.: Bulk density and changes depending on the moisture with different mechanical

and humus soil composition, Soil science and agrochemistry, XXVI, 3–4, Sofia, 1991.30

501





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Douville, H., Royer, J.-F., and Mahfouf, J.-F.: A new snow parameterisation for the Meteo-Franceclimate model. Part I: Validation in stand-alone experiments, Climate Dyn., 12, 21–35, 1995.

Dumenil, E. and Todini, L.: A rainfall-runoff scheme for use in the Hamburg climate model, in:Advances in Theoretical Hydrology, a tribute to James Dooge, edited by: O’Kane, J. P., page462, McGraw-Hill, New York, 1992.5

Etchevers, P., Golaz, C., and Habets, F.: Simulation of the water budget and the riverflows ofthe Rhone basin from 1981 to 1994, J. Hydrol., 244, 60–85, 2000.

Golaz, C., Etchevers, P., Habets, F., Ledoux, E., and Noilhan, J.: Comparison of two hydrologi-cal simulations of the Rhone basin, Phys. Chem. Earth, 26(5–6), 461–466, 2001.

Habets, F., Noilhan, J., Golaz, C., Goutorbe, J. P., Lacarrere, P., Leblois, E., Ledoux, E., Martin,10

E., Ottle, C., and Vidal-Madjar, D.: The ISBA surface scheme in a macroscale hydrologicalmodel applied to the Hapex-Mobilhy area. Part I : model and data base, J. Hydrol., 217,75–96, 1999a.

Habets, F., Noilhan, J., Golaz, C., Goutorbe, J. P., Lacarrere, P., Leblois, E., Ledoux, E., Martin,E., Ottle, C., and Vidal-Madjar, D.: The ISBA surface scheme in a macroscale hydrological15

model applied to the Hapex-Mobilhy area. Part II: simulation of streamflows and annual waterbudget, J. Hydrol., 217, 75–96, 1999b.

Hershkovich, E., Stephanov, I., Ganeva, B., et al.: Agro-climatical atlas of Bulgaria, Mean Direc-tion of Hydrology and Meteorology – Bulgarian Academy of Sciences, Institute of Hydrologyand Meteorology, Institute of Cartography, Sofia, 1982.20

Hottel, H. C.: A Simple Model for Estimating the Transmittance of Direct Solar RadiationThrough Clear Atmospheres, Solar Energy, 18, 129, 1976.

Kalinova, M.: Unconfined underground waters, in: Geography of Bulgaria – Physical Geog-raphy, 1982: Natural Conditions and Resources, Bulgarian Academy of Sciences, p. 259,Sofia, 1982.25

Kasten, F. and Czeplak, G.: Solar and terrestrial radiation dependent on the amount and typeof cloud, Solar Energy, 24, 177–189, 1979.

Ledoux, E., Girard, G., and Marsily, G.: Spatially distributed modelling: conceptual approach,coupling surface water and groundwater, in: Unsaturated Flow in Hydrologic Modelling The-ory and Practice, edited by: Morel-Seytoux, H. J., by Kluwer Academic Publishers, 1989.30

Machkova, M. and Dimitrov, D.: Book of reference for the quantitative characteristics of the un-derground waters for the period 1980–1996, National Institute of Meteorology and Hydrology,Sofia, 1999.

502





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Morel, S.: Modelisation distribuee du bilan hydrique a l’echelle regionale: application au bassinAdour Garonne, Doctoral Thesis, Toulouse 3 University, 2003.

Nash, J. E. and Sutcliffe, J. V.: Riverflow forecasting through conceptual models, 1, a discutionof principles, J. Hydrol., 10(3), 282–290, 1970.

Ninov, N.: Soils in the mountain area in: Geography of Bulgaria – Physical Geography, 1982:5

Natural Conditions and Resources, Bulgarian Academy of Sciences, 380–388, Sofia, 1982.Noilhan, J. and Lacarrere, P.: GCM gridscale evaporation from mesoscale modelling, J. Cli-

mate, 8(2), 206–223, 1995.Noilhan, J. and Planton, S.: A simple parameterization of land surface processes for meteoro-

logical models, Mon. Wea. Rev., 117, 536–549, 1989.10

Richards, L. A. and Wadleigh, C. H.: Soil water and plant growth, in: Soil Physical Conditionsand Plant Growth, edited by: Shaw, B. T., American Society of Agronomy Series MonographsVolume II, pp. 74–251, Academic Press, New York, 1952.

Rousset, F., Habets, F., Gomez, E., Le Moigne, P., Morel, S., Noilhan, J., and Ledoux, E.: Hy-drometeorological modeling of the Seine basin using the SAFRAN-ISBA-MODCOU system,15

J. Geophys. Res., 109, D14105, doi:10.1029/2003JD004403, 2004.Staley, D. and Jurica, G.: Effective atmospheric emissivity under clear skies, J. Appl. Meteor.,

11, 349–356, 1972.Troshanov, N.: Hidden discharge evaluation of the North Rhodopes karstic water to the Upper

Thracian lowland, Engineering Geology and Hydrogeology, 22, 10–27, Sofia, 1992.20

Vekilska, B. and Kalinova, M.: Snow pack in the West Stara Planina Mountain and its impacton the riverflow, Problems of the Geography of P. R. of Bulgaria 5, Sofia, Nauka i Izkustvo,1978.

Vekilska, B.: Radiation and heat budget in: Geography of Bulgaria – Physical geography, Nat-ural conditions and resources, 1982, pp 169–171, Bulgarian Academy of Sciences, Sofia,25

1982.Yordanova, M.: Genesis and monthly partition of the riverflow in the East Rhodopy Mountain – a

result from the complex effect of the physical and geographical factors, Thesis, GeographicalInstitute, Bulgarian Academy of Sciences, 1978.

Zyapkov, L.: Water budget of the river basins, in: Geography of Bulgaria – Physical Geography,30

1982: Natural Conditions and Resources, p 327, Bulgarian Academy of Sciences, Sofia,1982.

503





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Table 1. Sources of meteorological data, used for the preparation of model input database,following the station types and the time step of observation. Observers using traditional mea-surement instruments, except those for global solar radiation, make all the measures. Aurelhy(Benichou and Le Breton, 1987) and Bluepack interpolation packages were available at Meteo-France/CNRM.

Data type Unit Stations type/amount Time step of observation Used soft-ware ormethodfor spatialinterpolation

Synoptic Climatological Synoptic Climatological

Precipitation [mm] 16 55+120rain gauges

At 2, 8, 14, 20 hUTC6-h step

At 7 h LT – dailysum

Bluepack

Air Temper-ature at 2 m

[◦C] 16 55 3-h step At 7, 14 and 21h LT

Aurelhy orBluepack

Wind velo-city

[ms−1] 16 55 3-h step At 7, 14 and 21h LT

Bluepack

Global Solarradiation

[Wm−2] 2 stations – Chirpan and Rozhen 1 h sum Function

Sunshineratio

[1/10] 15 hourly Function

Cloudiness [1/10] 16 55 3-h step At 7, 14 and 21h LT

Bluepack

Relative hu-midity

[%] 16 55 3-h step At 7, 14 and 21h LT

Bluepack

504





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Table 2. Data collected for the validation of the modelling results: streamflow discharge wascomputed as daily average value, using hourly recordings or staff gage level observations andrating curves; dam’s inflow and release from Northern part of the Maritsa basin were obtainedas 10 day accumulated values, while from the southern reservoirs they were more detailed –with daily step; soil moisture was measured as difference before and after drying soil sam-ples extracted 3 times a month; snow water equivalent (SWE) was measured at five stationsmaintained by NEK.

Data type Unit Amount ofstations

Type of stations Data time step

Streamflow Discharge [m3/s] 56 River gages DailyDam inflow and release [m3/s] 12 Reservoir’s budgets Daily and 10 day

averagesSoil moisture in the 100 cmcolumn

[mm] 10 Agro-meteorological sta-tions – samples takenthe 7th, 17th and 27thday of the month

3 times monthly

Snow depth [cm] 55+120 Climatological + 120rain gauges

Daily

SWE [mm] 5 Climatological DailyPan evaporation [mm] 5 Climatological Daily

505





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Table 3. Average and extreme values of the five parameters for the reservoirs representing theunsaturated zone.

Parameter Average Maximum Minimum

α 0.75 1.0 0.05h1max 60 [mm] 300 5h2max 90 [mm] 300 1C1 4.0×10−3 4.0×10−2 1.0×10−4

C2 2.5×10−3 0.13 1.0×10−4

506





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Table 4. Comparison statistics of series of simulated against measured daily streamflow dis-charge for the main river gages on Maritsa and Tundzha rivers and four not perturbed by humanactivity watersheds: Varbitza, Krumovitza, Chepelarska and Mochuritza. Imposed streamflowtakes into account streamflow stored in or released from dam reservoirs, taking water from theriver bed and flow redirections. Qsim/Qobs is the simulated versus observed discharge ratio,E the daily efficiency, and R2 the coefficient of determination.

Calibration year withimposed streamflow(1995/1996)

Calibration year with-out imposed streamflow(1995/1996)

Validation year withoutimposed streamflow(1996/1997)

River,Gage station

Basinsurf.(km2)

Avg. alti-tude (m)

Qsim/Qobs

E R2 Qsim/Qobs

E R2 Qsim/Qobs

E R2

Maritsa,Plovdiv

7926 915 1.10 0.83 0.93 0.94 0.29 0.72 0.87 0.37 0.67

Maritsa,Svilengrad

20840 582 1.04 0.84 0.93 0.96 0.84 0.94 0.91 0.70 0.85

Tundzha,Elhovo

5551 475 1.00 0.64 0.82 1.16 0.47 0.79 0.75 0.54 0.76

Maritsa,whole basin

36255 1.16 0.75 0.95 1.02 0.65 0.88 1.10 0.47 0.86

Varbitza,Djebel

1149 584 0.89 0.77 0.88 0.89 0.77 0.88 0.96 0.78 0.89

Krumovitza,Krumovgrad

468 494 1.10 0.71 0.84 1.10 0.71 0.84 1.31 0.43 0.85

Chepelarska,Narechen

393 1356 1.09 0.72 0.87 1.09 0.72 0.87 0.95 0.59 0.77

Mochuritza,Vodenichene

1110 259 1.38 0.50 0.78 1.38 0.50 0.78 1.06 0.70 0.84

Average 1.09 0.72 0.87 1.07 0.62 0.84 0.99 0.57 0.81

507





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Table 5. Parameters and modelling results for the observed soil-crop profiles. The secondand third columns show the type of crops observed in the sites and the prescribed vegetationtypes used by the model for the corresponding grid cell. Next two columns show the fieldcapacity (Wfc) and wilting point (Wwilt) (values given for 1 m column depth), as observed andas prescribed in the model. Coefficients of determination of the comparison between simulatedversus observed series of normalized soil moisture are given in the last two columns: the firstone shows the statistics when all observed crop profiles of the site are considered, while thelast column shows the results obtained using only wheat or barley profiles (winter cereals).Plovdiv, Rajevo Konare and Ivailo sites are considered as influenced by irrigation.

LocationSoil type

Observed crop types Prescribed vege-tation types ag-gregation by orderof importance inthe grid cell

Wfc [mm] Wwilt [mm] Simulationresults forthe period95/97

Model Measure Model Measure Avg.R2

W&BR2

SadievoLeached Cinnamonic For-est soil (Alfisol)

BarleyCorn Grape

Grassland +Crops (2)

285 349 197 200 0.75 0.81

Liubenova MahalaSmolnitsa (Vertisol)

Barley Sunflower Crops (3) 332 404 244 257 0.68 0.81

Lucerne 377 231ChirpanSmolnitsa (Vertisol)

Wheat Corn Cotton Crops (3) 337 428 249 250 0.54 0.86

KazanlakDelluvial-meadow (Luvisol)

Wheat Rose Grassland 212 336 129 144 0.55 0.83

Mint 336 148HaskovoLeached Smolnitsa (EutricVertisol)

Wheat SunflowerGrape

Crops(3, 4, 2)

315 403 226 256 0.80 0.76

LiubimetsCinnamonic Forest soil (Al-fisol)

Wheat Sunflower Crops (3) 289 382 200 212 0.78 0.77

Grape 367 125PlovdivAlluvial-meadow (Fluvisol)

Wheat Corn Crops (3, 4) 245 303 158 175 0.62 0.62

Apple 389 175Rajevo KonareAlluvial-meadow (Fluvisol)

Wheat Sunflower To-bacco

Crops(3, 4, 2)

329 289 241 85 0.32 0.64

IvailoCinnamonic Forest soil (Al-fisol) – Podzols

Wheat Corn Apple Crops(3, 4, 2)

235 268 149 141 0.74 0.71

YambolSmolnitsa (Vertisol)

Wheat Corn Lucerne Crops(3, 4, 2)

337 386 249 207 0.73 0.76

508





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Table 6. Annual water budget for some main gauge stations and four not anthropized water-sheds: Ptot – total precipitation [mm]; Psnw – snow precipitation [mm]; Etot – total evaporation[mm]; Qtot – total runoff [mm]; Dw – evolution of soil water storage [mm]; Eg – evaporation fromthe bare soil [mm]; Er – plant interception evaporation [mm]; Etr – plant transpiration [mm]; Es– sublimation/evaporation at the snow surface [mm]; ISBA drainage D and surface runoff Qr[mm]; Storage in the snow pack is neglected for all the watersheds as the simulation ends on30 September, when snow pack rarely exists.

River, (sub-basin) 1995–1996 Surface Ptot Psnw Etot Qtot Dw Eg Er Etr Es D Qr

Maritsa, Edirne (whole basin) 36 255 621 163 433 163 25 177 90 139 26 125 38Maritsa, Svilengrad 21 379 598 174 430 140 28 174 91 136 29 105 35Maritsa, Plovdiv 8077 623 211 423 164 36 160 106 116 41 125 39Tundzha, Elhovo 5549 651 128 484 120 48 193 101 166 24 83 37Chepelarska, Narechen 412 765 317 460 283 21 149 122 120 70 223 61Varbitza, Djebel 1144 751 162 381 359 11 167 85 114 15 293 66Krumovitza, Krumovgrad 531 811 185 368 445 −2 160 77 113 18 365 80Mochuritza, Vodenichene 1190 680 91 470 140 70 187 84 185 14 97 42

River, (sub-basin) 1996–1997 Surface Ptot Psnw Etot Qtot Dw Eg Er Etr Es D Qr

Maritsa, Edirne (whole basin) 36 255 652 105 513 171 −32 206 97 193 16 126 45Maritsa, Svilengrad 21 379 603 106 510 125 −32 207 92 193 18 87 38Maritsa, Plovdiv 8077 595 149 519 112 −36 203 102 184 31 75 37Tundzha, Elhovo 5549 669 94 572 154 −58 211 109 237 16 109 45Chepelarska, Narechen 412 900 316 583 341 −25 199 141 170 74 265 76Varbitza, Djebel 1144 894 83 457 473 −36 193 100 157 6 371 102Krumovitza, Krumovgrad 531 938 152 404 566 −33 170 86 138 10 453 113Mochuritza, Vodenichene 1190 654 62 594 139 −79 215 106 266 7 98 42

509





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Fig. 1. Map of the Maritsa basin in Bulgaria. The red boundary line represents the politicalborders between Bulgaria and Turkey, Bulgaria and Greece, Turkey and Greece. The basinborder is shown with black line. The modelling area goes down to the town of Edirne in Turkey,where the watersheds of the Arda (from West) and Tundzha rivers (from North) reach the mainriver course.

510





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Fig. 2. Surface hydrological network of the Maritsa river system: the river meshes are in darkcolour, grey colour represents the aquifer area; dams are shown with dark box; the largersprings with diamond and river gauges with red circles; karstic areas in Maritsa river basin areshown too.

511





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

-40

-30

-20

-10

0

10

20

30

11/1

995

12/1

995

01/1

996

02/1

996

03/1

996

04/1

996

05/1

996

06/1

996

07/1

996

08/1

996

09/1

996

10/1

996

[mm

]

Transfer

Channels-In

Channels-Out

Irrigation

Dam-Release

Dam-Inflow-8

-6

-4

-2

0

2

4

6

11/1

995

12/1

995

01/1

996

02/1

996

03/1

996

04/1

996

05/1

996

06/1

996

07/1

996

08/1

996

09/1

996

10/1

996

[mm

]

Totalinfluence

ba

Fig. 3. (a) Monthly budget of the anthropogenic influence on natural riverflow for the entirebasin in Bulgaria in [mm]. “Dam-Inflow” – the dam reservoirs inflow, “Dam-Release” – the damreservoirs outflow, “Channels-In” – added water into the riverflow from within Maritsa basin,“Channels-Out” – redirected part of the riverflow within Maritsa basin, “Transfer” – additionalwater from outside the Maritsa basin, “Irrigation” – water used for irrigation purposes. (b)Overall effect of the anthropogenic influence on natural riverflow [mm].

512





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Fig. 4. Scheme of the ISBA – MODCOU coupled model with the 2 additional reservoirs for thedrainage, representing the unsaturated layer: H – sensible heat flux, LE – evaporation (latentheat) flux, G – ground heat flux, D – ISBA drainage, Qr – ISBA surface runoff, Qd – finaldrainage.

513





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Fig. 5. Map of the dominant vegetation types: forests are dominant in the mountain ranges;crops, and especially crop #3 and #4 – interpreted as Mediterranean cereal, are prevalent inthe valley; rocks are appearing close to mountain peaks.

514





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Fig. 6. Annually averaged atmospheric fields for 1996 (left) and 1997 (right). (a) and (b) Totalprecipitations [mm]; (c) and (d) Snow precipitations [mm]; (e) and (f) Air Temperature at 2 m[◦C]; (g) and (h) Global Radiation [Wm−2].

515





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

0

10

20

30

40

50

60

70

80

90

100

05/0

8/19

95

05/0

9/19

95

05/1

0/19

95

05/1

1/19

95

05/1

2/19

95

05/0

1/19

96

05/0

2/19

96

05/0

3/19

96

05/0

4/19

96

05/0

5/19

96

05/0

6/19

96

05/0

7/19

96

05/0

8/19

96

05/0

9/19

96

05/1

0/19

96

05/1

1/19

96

05/1

2/19

96

05/0

1/19

97

05/0

2/19

97

05/0

3/19

97

05/0

4/19

97

05/0

5/19

97

05/0

6/19

97

05/0

7/19

97

05/0

8/19

97

05/0

9/19

97

10-d

ay a

ccum

ulat

ed e

vapo

ratio

n [m

m]

average of 5 measuring sites

average of 5 simulated grid cellsR2=0.87RMSE=5.3

Fig. 7. Series of 10-day accumulated Penman evaporation computed with the model forcingand observed 10 day accumulated pan evaporation [mm]. Each point value is averaged fromthe observations of five stations or from the corresponding five grid cells values.

516





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

a b

c d

e f

g h

0

50

100

150

200

250

300

10 12 02 04 06 08 10 12 02 04 06 08

m3/s simulatedMaritsa_Plovdiv

0

100

200

300

400

500

600

10 12 02 04 06 08 10 12 02 04 06 08

m3/s simulatedMaritsa_Svilengrad

0

500

1000

1500

2000

10 12 02 04 06 08 10 12 02 04 06 08

m3/s simulatedMaritsa_Edirne

0

50

100

150

200

250

300

350

10 12 02 04 06 08 10 12 02 04 06 08

m3/s simulatedTundzha_Elhovo

0

100

200

300

400

500

600

700

800

900

1000

10 12 02 04 06 08 10 12 02 04 06 08

m3/s simulatedVarbitza_Djebel

0

50

100

150

200

250

10 12 02 04 06 08 10 12 02 04 06 08

m3/s simulatedKrumovitza_Krumovgrad

0

5

10

15

20

25

30

35

40

45

10 12 02 04 06 08 10 12 02 04 06 08

m3/s simulatedChepelarska_Narechen

0

5

10

15

20

25

30

35

40

45

50

10 12 02 04 06 08 10 12 02 04 06 08

m3/s simulatedMochuritza

Fig. 8. Comparison of the observed (solid line) and simulated (dashed line) hydrographs formain Maritsa and Tundzha subbasins (a, b, c and d) and four watersheds not disturbed byhuman activities (e, f, g and h) for the calibration (1995/1996) and validation (1995/1996) hy-drological years. 517





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Av

a b

c d

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100

average measured snow depth [cm]

grid

cel

l ave

rage

sno

w d

epth

[cm

] R2=0.44RMSE=8.92

0

20

40

60

80

100

120

01/1

0/96

31/1

0/96

30/1

1/96

30/1

2/96

29/0

1/97

28/0

2/97

30/0

3/97

29/0

4/97

29/0

5/97

snow

dep

th [c

m]

OBSMODEL

RMSE=13.2R2=0.89E=0.7

0

20

40

60

80

100

12001

/11/

95

01/1

2/95

31/1

2/95

30/0

1/96

29/0

2/96

30/0

3/96

29/0

4/96

snow

dep

th [c

m]

OBS

MODELRMSE=5,1R2=0.94E=0.9

0

5

10

15

20

25

30

35

01/1

1/95

01/1

2/95

31/1

2/95

30/0

1/96

29/0

2/96

30/0

3/96

29/0

4/96

SW

E [m

m]

OBSMODEL

Fig. 9. Snow depth and snow water equivalent (SWE) comparisons: Comparison of the av-eraged on 20 points of the observed and simulated snow depth [cm] – (a) winter 1995/1996;(b) winter 1996/1997; (c) Comparison of the averaged on 5 sites (SWE) observed and simu-lated [mm] (d) Scatter plot of the average simulated snow depth [cm] compared to the averageobserved snow depth for the whole observation network (174 gauges).

518





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

Fig. 10. Annually accumulated fields for the water budget components to the left for 1995/1996and to the right for 1996/1997: (a)–(b) Total evaporation [mm]; (c)–(d) Runoff [mm].

519





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

0

20

40

60

80

100

120

10/1

995

12/1

995

02/1

996

04/1

996

06/1

996

08/1

996

10/1

996

12/1

996

02/1

997

04/1

997

06/1

997

08/1

997

mm

Rain [mm] Snowfall [mm]Evaporation [mm] Available for the runoff [mm]

a

0

10

20

30

40

50

60

70

80

90

100

10/1

995

12/1

995

02/1

996

04/1

996

06/1

996

08/1

996

10/1

996

12/1

996

02/1

997

04/1

997

06/1

997

08/1

997

mm

LETR

LER

LEG

LES

b

0

50

100

150

200

10/1

995

12/1

995

02/1

996

04/1

996

06/1

996

08/1

996

10/1

996

12/1

996

02/1

997

04/1

997

06/1

997

08/1

997

mm

Snow reservoir

Unsaturatedzone reservoirSoil reservoir

c

Fig. 11. Monthly evolution of the water budget of the whole basin: (a) Main water budgetcomponents – Rain and snow precipitations (stacked areas), Total evaporation and Runoff; (b)Evaporation components: Bare ground evaporation (LEG), Plant transpiration (LETR), Inter-cepted water evaporation (LER) and evaporation/sublimation from snow surface (LES) (stackedcolumns representation except LES); (c) Main water storage reservoirs’ evolution: soil reser-voir, unsaturated zone reservoir and snow reservoir (stack column representation). The unit forall the variables is mm/year. 520





HESSD4, 475–521, 2007


E. Artinyan et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



EGU

a

b

0,0E+00

5,0E+03

1,0E+04

1,5E+04

2,0E+04

2,5E+04

3,0E+04

0,0E+00 5,0E+03 1,0E+04 1,5E+04 2,0E+04 2,5E+04 3,0E+04

basin range modelling results for 2 years [Wm-2]

clim

atol

ogic

al, c

ount

ry r

ange

val

ues

[Wm

-2] Rn year

LE year

Rn August

H year

LE January

LE June

Rn December

0

20

40

60

80

100

120

140

10/1

995

12/1

995

02/1

996

04/1

996

06/1

996

08/1

996

10/1

996

12/1

996

02/1

997

04/1

997

06/1

997

08/1

997

[Wm

-2]

LE G

Rn H

Fig. 12. (a) Monthly values of the energy budget for the whole basin: Rn – Net radiationflux [Wm−2], LE – Latent heat flux [Wm−2], H – Sensible heat flux [Wm−2], G – Ground heatflux [Wm−2]; (b) Basin-range model energy budget components compared to climatologicalcountry-range energy budget components.

521





Modelling the water budget and the riverflows of the Maritsa basin in Bulgaria

Documents